VOOZH about

URL: https://www.phoronix.com/review/arm_cortexa15_opts/2

⇱ ARM Cortex-A15 Compiler Optimizations - Phoronix


👁 Phoronix

ARM Cortex-A15 Compiler Optimizations

Written by Michael Larabel in Software on 28 February 2013 at 10:28 AM EST. Page 2 of 3. Add A Comment.

For GraphicsMagick, -O0 is obviously the slowest with the compiler not applying any optimizations. The -Os performance was close to -O1. For some reason, GraphicsMagick when building with GCC 4.7.2 on ARM when setting the CFLAGS/CXXFLAGS to -O3 would only be honored as -O2. Using -Ofast didn't yield any greater gains.

-O2/-O3 produced the fastest binaries for Himeno, which were more than twice as fast as -O0 without any optimizations. Passing -Ofast actually led to a minor setback in performance compared to -O2.